Vol.2 No.3 2009

Research paper : Creating non-volatile electronics with spintronics technology (S. Yuasa et al.)−204−Synthesiology - English edition Vol.2 No.3 (2009) of the magnetoresistance effect.Yoshisige SuzukiCompleted the Tsukuba University Science and Engineering Research Department Master’s program in March 1984. Joined the Electrotechnical Laboratory (now AIST) in April 1984. PhD., Engineering. After serving as Director of the Electronics Division Spintronics Research Group, he has been serving both as Professor in the Osaka University Graduate School Engineering Science and as Guest Researcher at AIST since March 2004. He has been engaged in basic and applied research in spintronics. He has received the Prime Minister Award (Outstanding Contribution to Cooperation by Industry, Academia and Government) (2008), Applied Physics Society Best Paper Award (2005, 2007), and two others. In this paper, he was responsible for development of measurement and evaluation methods for spin torque magnetization reversal and spin dynamics.Koji AndoCompleted the Tokyo Institute of Technology Graduate School of Science and Technology doctoral program in March 1975. Joined the Electrotechnical Laboratory (now AIST) in April 1975. PhD., Engineering. Became Assistant Laboratory Director of the Electronics Research Department in November 1984. He has been engaged in basic and applied research in spintronics, with emphasis on the magneto-optical effect and magnetic semiconductors. He has received the Applied Physics Society Fellow Award (2008), the Japan Society of Applied Magnetics Performance Award (2007), and the Applied Physics Society Best Paper Award (2005). In this paper, he was responsible for NEDO Spintronics Non-volatile Device Project, and joint research with Toshiba, as well as organization of the overall concept.Discussion with Reviewers1 Achieving the breakthroughQuestion and comment (Naoto Kobayashi, Center for Research Strategy, Waseda University)I think this R & D is a rare example where the research results, through the two major breakthroughs of “ 1) the giant TMR effect achieved using MgO” and “ 2) the realization of CoFeB/MgO/CoFeB structure” are linked to the outcome of development of a product that is actually sold in the market in a short period of time. From this paper we can understand the most part why these respective breakthroughs could be achieved, but could we have an explanation of the process of selecting those processes and materials (including the reasons for excluding other materials and processes, etc., and, if necessary, the research and development system). Also, if there was an effective serendipity or other such factors, please also describe them.Answer (Shinji Yuasa)Half of the success factors were a matter of achieving set goals, and the other half involved good luck (serendipity). Theoretically, there are a number of promising crystalline tunnel barrier materials other than MgO. The first thing to consider is the problem of which theoretical prediction is correct (which prediction should we bet on). Furthermore, to reach a practical stage means the satisfaction of various requirements such as (i) fabricating thin films of nanometer order thickness without opening up pinholes, (ii) prevention of reaction and atomic diffusion at the interface with the metal electrode material, (iii) crystallization in low temperature layer formation, (iv) a sufficiently high breakdown voltage, and (v) product-level reliability. Based on the experience and knowledge we have accumulated and as the result of considering what material system would be the most promising from many angles, we came to the conclusion that using anything other than crystalline MgO would present difficulties even before we began any experimentation. However, that did not at all mean that there would be no problems if we chose to go with MgO. The most troubling problem before beginning the research and development was production process compatibility, but that problem was solved by the development of the CoFeB/MgO/CoFeB structure in joint research with Canon ANELVA Corporation. Serendipity was large among the success factors in forming the CoFeB/MgO/CoFeB structure, I believe, even with the active work done by the excellent engineers of the Canon ANELVA Corporation. Also, the inability to judge ultimate reliability at the product level until the final stage of research and development is reached is a problem, but we had the feeling from the initial stage of research and development that the reliability of MgO was high. That was not a conclusion based on theoretical consideration, but an intuition that can only be felt by people working in the lab.2 Reduction of power consumptionQuestion and comment (Naoto Kobayashi)You have described that the development of non-volatile memory is linked to the realization of ultimate green IT devices. While practical non-volatile memory will greatly reduce the power needed to operate memories, I think that no overall reduction in power consumption can be expected unless the non-volatile CPU is realized as described in Fig. 1. Thus I have two questions: 1) As far as we know now, what are the approximate relative proportions of the power consumption decrease that can be achieved by implementing the spin RAM and that achievable by implementing the future non-volatile CPU? and 2) I believe that spin FETs is essential to implementation of future non-volatile CPUs. What do you think of the prospects for that field of research and development?Answer (Shinji Yuasa)1) Not much reduction in power consumption can be expected from simply replacing DRAM and SRAM with MRAM and spin RAM. As you point out, a radical reduction in power consumption would require achieving non-volatility for both memory and the CPU together as a set. While non-volatile DRAM and SRAM serve as the first stage, the merits of that in itself include (i) an opening up of the near-future SRAM and DRAM scaling limits and (ii) higher integration scale and lower cost achieved by on-chip system LSI memory consisting of spin RAM only.2) While switching devices that have a non-volatile memory function such as spin FETs are ideal, one opinion that non-volatile logic circuits can also be designed by combining existing memory devices (MTJ devices and ferro-electric memory devices, etc.) and CMOS. In any event, achieving normally-off equipment with a non-volatile CPU is a grand plan that will require considerable time and investment, and so is expected to need R & D on a 20-year scale.Question and comment (Kazuhito Ohmaki, Department of Information and Computer Sciences, Toyo University)Particularly in the Introduction, TMR is proposed as a key technology for normally-off computers. With today’s computers, it may look like the computer is idle between key inputs, but in the meantime, the computer is busy monitoring communication lines,


page 31